There is a significant regression of startup time of Netbeans IDE when GTK L&F is used instead of Metal L&F on Linux. Below are numbers measured under the following conditions:

* Dell Precision 220, PIII 800MHz, 512MB RAM, Red Hat 9, Gnome
* Blueprint theme (JDS default) and Bluecurve theme (RH9 default)
* jdk 1.5.0-beta2-b51
* netbeans trunk build 200405171800
* times measured manually using a stop watch
(from hitting enter in console till the moment the IDE's main window is fully painted)
* the first start in each of the combinations is creating a new userdir, the next two starts are using this created userdir

Cold starts (each time measured after a fresh reboot of the machine):

[ L & F ] - [ blueprint theme ] - [ bluecurve theme ]
Metal L&F - 27.5s, 25.2s, 25.0s - 29.7s, 24.7s, 24.6s
GTK L&F - 29.3s, 27.2s, 27.7s - 31.2s, 27.0s, 26.8s
increase = 6.5%, 7.9%, 10.8% - 5.0%, 9.3%, 8.9%

Warm starts (first a warmup start is done and then next starts are measured without rebooting):

[ L & F ] - [ blueprint theme ] - [ bluecurve theme ]
Metal L&F - 18.4s, 15.8s, 15.5s - 17.4s, 15.2s, 15.1s
GTK L&F - 20.6s, 17.8s, 17.6s - 20.2s, 18.0s, 17.2s
increase = 12.0%, 12.7%, 13.5% - 16.1%, 18.4%, 13.9%

###@###.### 2004-05-18
###@###.### 2004-05-18